Re: ACPI throttling changes

From: Alexandre (Alex.Kovalenko_at_verizon.net)
Date: 12/13/03

  • Next message: Robert Watson: "Re: [RC1] Login not possible"
    Date: Fri, 12 Dec 2003 19:23:14 -0500
    To: Nate Lawson <nate@root.org>
    
    

    On Tue, 9 Dec 2003 17:57:53 -0800 (PST)
    Nate Lawson <nate@root.org> wrote:

    > I'm working on a shared CPU frequency control driver. One step is to
    > remove some of the autonomy of the throttling portion of acpi_cpu.
    > Please test this patch if you have a machine which supports throttling.
    > With this patch, throttling can be changed by doing:
    >
    > sysctl hw.acpi.cpu.current_speed=X
    >
    > where X is some number between 1 and hw.acpi.cpu.max_speed. It is no
    > longer driven by AC line transitions. Run a CPU benchmark like this one
    > to make sure the throttling transition still works with this patch.
    >
    > dd if=/dev/zero bs=1m count=500 | md5
    >
    > This is part of a larger work. Don't worry, it won't be committed until
    > general CPU frequency control is done so no loss of functionality will be
    > committed.
    >
    > -Nate

    I do not know if it's still of any use (I just caught up on my reading),
    but for whatever it worth, results are below. To be fair results are a
    little bit surprising to myself as I "feel" better response with 'economy'
    set to 4 then with 'economy' set to 2, but of course this in unscientific.

    I'm running -CURRENT from yesterday + this patch + my patch for picking PCI
    IRQ, but the latter shoud not impact performance in any way shape or form.

    Please, let me know if there is any additional information I can provide or
    if there any other things you want me to test here.

    Trottling to 1
    hw.acpi.cpu.current_speed: 8 -> 1
    104857600 bytes transferred in 8.961181 secs (11701315 bytes/sec)
        9.17s real 6.21s user 2.70s system
    Trottling to 2
    hw.acpi.cpu.current_speed: 1 -> 2
    104857600 bytes transferred in 8.987054 secs (11667628 bytes/sec)
        9.05s real 6.20s user 2.75s system
    Trottling to 3
    hw.acpi.cpu.current_speed: 2 -> 3
    104857600 bytes transferred in 8.983341 secs (11672450 bytes/sec)
        9.05s real 6.08s user 2.83s system
    Trottling to 4
    hw.acpi.cpu.current_speed: 3 -> 4
    104857600 bytes transferred in 9.006142 secs (11642899 bytes/sec)
        9.07s real 6.17s user 2.78s system
    Trottling to 5
    hw.acpi.cpu.current_speed: 4 -> 5
    104857600 bytes transferred in 8.984664 secs (11670731 bytes/sec)
        9.05s real 6.25s user 2.68s system
    Trottling to 6
    hw.acpi.cpu.current_speed: 5 -> 6
    104857600 bytes transferred in 8.986324 secs (11668575 bytes/sec)
        9.05s real 6.12s user 2.82s system
    Trottling to 7
    hw.acpi.cpu.current_speed: 6 -> 7
    104857600 bytes transferred in 8.989106 secs (11664964 bytes/sec)
        9.06s real 6.23s user 2.72s system
    Trottling to 8
    hw.acpi.cpu.current_speed: 7 -> 8
    104857600 bytes transferred in 7.827533 secs (13395996 bytes/sec)
        7.89s real 5.36s user 2.43s system

    _______________________________________________
    freebsd-current@freebsd.org mailing list
    http://lists.freebsd.org/mailman/listinfo/freebsd-current
    To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"


  • Next message: Robert Watson: "Re: [RC1] Login not possible"

    Relevant Pages

    • Re: [PATCH] AMD Thermal Interrupt Support
      ... with my formerly submitted patch. ... to the throttling message -- not just the device. ... initialization consists of 3 parts: ... a thermal interrupt is delivered. ...
      (Linux-Kernel)
    • Re: [PATCH] AMD Thermal Interrupt Support
      ... I'll add a robust description with the next revision of the patch. ... My buddy at Google suggested that I might as well fix this while I'm ... the throttling MCEs which you might naturally expect to see ... if you're expecting to be warned at 50C that your CPU has ...
      (Linux-Kernel)
    • Re: [PATCH 00/15] per device dirty throttling -v6
      ... I put in quite a few comments, and added an patch to do per task dirty ... code for per-task throttling. ... So if there is one writer, it now gets only half the space that it ...
      (Linux-Kernel)
    • Re: [PATCH 00/15] per device dirty throttling -v6
      ... I put in quite a few comments, and added an patch to do per task dirty ... code for per-task throttling. ... So if there is one writer, it now gets only half the space that it ...
      (Linux-Kernel)
    • Re: [acpi-jp 2870] ACPI throttling changes
      ... > I'm working on a shared CPU frequency control driver. ... > Please test this patch if you have a machine which supports throttling. ... > to make sure the throttling transition still works with this patch. ...
      (freebsd-current)